Programmable Processor Design for Givens Rotations based Applications

Singular value decomposition on GPU using CUDA

“…Bondhugula et al [7] report only the time for the bidiagonalization of the matrix. Dickson et al [11] presented a programmable processor design suitable for SVD, but do not give any SVD results. Yamamoto et al [25] give the optimized algorithm only for large rectangular matrices.…”

“…Baker [2] described a parallel algorithm to compute the SVD of block circulant matrices on Cray-2. Dickson et al [11] designed a programmable processor for computing the Givens rotation using approximate rotation method. The processor can also be programmed for SVD computation.…”